Update Cratis.Chronicle to 16.36.3

[cratis-bot]

Summary

Cratis.Chronicle, Cratis.Chronicle.AspNetCore and Cratis.Chronicle.Testing are updated from 16.35.3 to 16.36.3, picking up 16.35.4, 16.35.5, 16.35.6, 16.35.7, 16.35.8, 16.36.0, 16.36.1, 16.36.2 and 16.36.3.

Added

• Chronicle 16.35.4: Every published Docker image variant (production, development, development-slim, workbench) is booted the way a consumer runs it before being pushed — the container must start, stay up, and listen on its port, or the publish fails with the container logs (Fix Chronicle server container images failing to start with Permission denied Chronicle#3446)
• Chronicle 16.36.0: CannotWaitForObserverCompletion names the case where completion cannot be established

Changed

• Cratis.Chronicle, Cratis.Chronicle.AspNetCore and Cratis.Chronicle.Testing are updated to 16.36.3
• Chronicle 16.36.0: Waiting for observer completion on an append that carries no observer surface now fails by name instead of reporting success, so an in-process spec can no longer assert that downstream work finished when none of it was started

Fixed

• Chronicle 16.35.6: Asking a MongoDB-backed event sequence for the next sequence number at or after a position now answers with the nearest qualifying event instead of the furthest, matching the SQL and in-memory implementations (GetNextSequenceNumberGreaterOrEqualThan returns the highest match on MongoDB, so failed-partition retry skips events Chronicle#3743)
• Chronicle 16.35.8: The image boot gate no longer fails every production publish: it now supplies the TLS and token-encryption certificates the production image requires, instead of starting it in a configuration it deliberately refuses
• Chronicle 16.36.1: Production Docker images publish again — none were published for 16.35.4 through 16.36.0, and the latest tag stayed on 16.35.3 (No production cratis/chronicle image has been published since 16.35.3 — six releases including 16.36.0 ship only the -development variants Chronicle#3770)
• Chronicle 16.36.1: Job state now round-trips correctly on SQL storage backends (SQLite, PostgreSQL and MSSQL) — a job's request payload was lost when its state was written and reading it back failed, leaving projection retirement silently incomplete. MongoDB was unaffected
• Chronicle 16.36.2: Publishing several merges at once no longer loses releases to a race on the next version number: runs queue behind each other instead of computing the same version in parallel
• Chronicle 16.36.3: Jobs, users, and applications can be read through generated gRPC clients without timestamp serialization failures. (Generated gRPC contracts cannot serialize DateTimeOffset responses Chronicle#3788)

Removed

• Chronicle 16.36.1: The pre-publish Docker boot check

[cratis-bot]

Verification gate

Step	Result
Build	failure
Specs	skipped
Bump	patch

⚠️ Requires human review before merging — verification failed.